A network-level sidewalk inventory method using mobile LiD AR and deep learning

摘要

Sidewalks are a critical infrastructure to facilitate essential daily trips for pedestrian and wheelchair users. The dependence on the infrastructure and the increasing demand from these users press public transportation agencies for cost-effective sidewalk maintenance and better Americans with Disabilities Act (ADA) compliance. Unfortunately, most of the agencies still rely on outdated sidewalk mapping data or manual survey results for their sidewalk management. In this study, a network-level sidewalk inventory method is proposed by efficiently segmenting the mobile light detection and ranging (LiDAR) data using a customized deep neural network, i.e., PointNet++, and followed by integrating a stripe-based sidewalk extraction algorithm. By extracting the sidewalk locations from the mobile LiDAR point cloud, the corresponding geometry features, e.g., width, grade, cross slope, etc., can be extracted for the ADA compliance and the overall condition assessment. The experimental test conducted on the entire State Route 9, Massachusetts has shown promising performance in terms of the accuracy for the sidewalk extraction (i.e., point-level intersect over union (IoU) value of 0.946) and the efficiency for network analysis of the ADA compliance (i.e., approximately 6.5 min/mile). A case study conducted in Columbus District in Boston, Massachusetts, demonstrates that the proposed method can not only successfully support transportation agencies with an accurate and efficient means for network-level sidewalk inventory, but also support wheelchair users with accurate and comprehensive sidewalk inventory information for better navigation and route planning.